Nest Site Selection during Colony Relocation in Yucatan Peninsula Populations of the Ponerine Ants Neoponera villosa (Hymenoptera: Formicidae)

In the Yucatan Peninsula, the ponerine ant Neoponera villosa nests almost exclusively in tank bromeliads, Aechmea bracteata. In this study, we aimed to determine the factors influencing nest site selection during nest relocation which is regularly promoted by hurricanes in this area. Using ants with and without previous experience of Ae. bracteata, we tested their preference for refuges consisting of Ae. bracteata leaves over two other bromeliads, Ae. bromeliifolia and Ananas comosus. We further evaluated bromeliad-associated traits that could influence nest site selection (form and size). Workers with and without previous contact with Ae. bracteata significantly preferred this species over others, suggesting the existence of an innate attraction to this bromeliad. However, preference was not influenced by previous contact with Ae. bracteata. Workers easily discriminated between shelters of Ae. bracteata and A. comosus, but not those of the closely related Ae. bromeliifolia. In marked contrast, ants discriminated between similar sized Ae. bracteata and Ae. bromeliifolia plants, suggesting that chemical cues and plant structure play an important role. Size was also significant as they selected the largest plant when provided two dissimilar Ae. bracteata plants. Nest site selection by N. villosa workers seems to depend on innate preferences but familiarization with plant stimuli is not excluded.

Neoponera villosa is a generalist arboreal predatory ant [18][19][20] with a wide geographical distribution, from Texas to Argentina [21]. This species occurs both in wet and dry forests [22] and is an opportunistic cavity breeder that nests in dead and live trees, and in bromeliads [17,23,24]. In the southern part of the Yucatan Peninsula, Mexico, N. villosa nests mainly in the epiphytic bromeliad Aechmea bracteata In the present study, we performed different experiments (two-choice bioassays) to determine how N. villosa workers select a nest site in the eventuality of nest relocation. Because rearing workers from egg to adult was not feasible, we took advantage of the fact that N. villosa nests almost exclusively in cavities of live trees in northern Yucatan where Ae. bracteata is rare, to investigate nest site selection of N. villosa workers without previous contact with this bromeliad. Our research addressed the following questions: (1) Do N. villosa workers have an innate preference for Ae. bracteata? (2) Is the preference modulated by the preimaginal or neonatal ant experience linked to the origin of the colony (workers with or without previous contact with Ae. bracteata)? (3) Are the recognition and localization of Ae. bracteata regulated by chemical stimuli? (4) Does Ae. bracteata size influence nest site selection?

Ant Collection and Identification
Ants in bromeliads were collected in the following five sites in the southern part of the Yucatan Peninsula: Ejido Blasillo (18 •  In the latter two sites, Ae. bracteata is rare or absent. Each epiphyte was cut off from the supporting branch, dismantled leaf-by-leaf in plastic bins coated with Fluon (Whitford GmbH, Diez, Germany), and all N. villosa ants and their brood were collected. Both ants from epiphytes and those from tree cavities were housed in plastic jars under laboratory conditions until bioassays (see below).
Neoponera villosa belongs to the neotropical species complex of N. foetida (L.), which includes 12 other species [56,57]. Due to their morphological similarity, N. villosa has been confused in the past with two other species with a wide distribution, N. inversa (Smith) and N. curvinodis (Forel). Until now, however, only N. villosa has been reported in the Yucatan Peninsula [58]. Nevertheless, in order to confirm ant identity and further support our comparisons, five workers nesting in bromeliads and five workers nesting in live trees were DNA extracted and barcoded as part of an independent study (Lachaud and Pérez-Lachaud, unpubl.). DNA extraction and amplification followed the protocol in [59], with a freezing step after initial incubation according to the recommendation of [60] for Hymenoptera. Sequences were edited using CodonCode v. 3.0.1 (CodonCode Corporation, Dedham, MA, USA) and uploaded to the Barcode of Life Database (BOLD, boldsystems.org). Voucher specimens were deposited in the Formicidae Collection of El Colegio de la Frontera Sur at Chetumal, Quintana Roo, Mexico (ECO-CH-F).

Nest Site Selection
As our study is focused on nest relocation, only workers were used for the two-choice tests implemented to evaluate nest site selection. This parallels nest relocation following disturbance or destruction of the old nest, whereby emigrations are organized by workers (scouts) that set out from the damaged nest to find a new home, thoroughly inspecting any candidate nest that they find [2]. As in various other ponerine species such as N. verenae (Forel) (referred to as Pachycondyla obscuricornis Emery [61]), N. apicalis (Latreille) [62], or Diacamma indicum (Santschi) [63,64], N. villosa uses a specific behavior called tandem running in which a single worker attracts a single recruit (or two in some occasions) and leads her towards the new nest site [17]. As in other Neoponera [61,62], such recruitment by N. villosa is exclusively used during nest relocation and never during foraging, see [18].

Experimental Setup
Ants from 35 colonies living in Ae. bracteata and 11 colonies (or parts of colonies) nesting in tree cavities were used in bioassays (Table S1). Observations were conducted from 26 January 2017 to 12 February 2018. Bioassays were carried out two weeks after field collection; during this period of acclimatization, and during bioassays, ants were held under natural illumination and at room temperature (26 ± 1 • C). Workers were randomly selected from the original nest and only evaluated once. Two-choice tests were conducted using transparent plastic jars (14 cm in diameter × 25 cm in height, 3 L vol.) into which the ants were deposited. Each jar was connected via a transparent plastic tube (1.5 cm in diameter × 20 cm in length) to an election chamber (bioassays with live plants: 45 × 30 × 60 cm plastic box; bioassays with parts of plants (leaves): 40 × 21 × 14 cm). Each election chamber included a glass tube (2 cm in diameter x 8 cm in length) filled with water and stuffed with cotton at one end. The ants were fed sliced apple pulp, which was placed in the election chamber for the duration of the bioassay.
The protocol for the observations followed that of [25]. Behavioral heterogeneity among colony members is common in insect societies and individual behavioral specialization during nest moving has been signaled, for example, in the model ant genus Temnothorax [65,66]; however, specialized Temnothorax workers are readily replaced in removal experiments showing organizational resilience of ant colonies [65]. As colony size in N. villosa varies greatly from one colony to another, and because it was not feasible to collect a sufficient number of complete colonies to perform the required number of replicates per bioassay, a fixed sample size of 20 ants per bioassay was used. Furthermore, this is a common procedure in experiments with large ants (see for example [36,37]). For each bioassay, two different refuges or "nest sites", consisting of tubular shelters to eliminate the influence of plant architecture, were placed in the election chamber. Subsequently, a group of 20 workers randomly obtained from those foraging and some brood were gently placed into the adjacent transparent jar. We carried out 21 to 30 replicates for each comparison and each replicate consisted of individuals from the same colony. Then, the experimental device was closed and set aside for 24 h, allowing the ants to install themselves in one shelter along with the brood (see [36]). The stimulus for the initiation of movement towards a potential nest (no shelter in the jar and artificial illumination) was constant across experiments and across replicates within experiments. This is a standard procedure used to trigger colony relocation in ants [67]. After 24 hours we evaluated the number of workers in any of the two refuge options ("nest sites") and those that remained in the jar or that were wandering or foraging.

Experiment One
To evaluate whether N. villosa nests in any available cavity or whether it prefers the refuge provided by Ae. bracteata, workers were presented with two tubular shelters (4 cm in diameter × 10 cm in length, with only one opening) made from the following: (a) the rolled leaves of Ae. bracteata (treatment) and (b) from a cardboard (control). Thirty replicates were performed with workers originating from colonies living in bromeliads and thirty with workers from colonies collected in tree cavities.

Experiment Two
To evaluate whether N. villosa workers are able to discriminate Ae. bracteata through chemical stimuli emitted by the plant, shelters made of leaves of two other species belonging also to the Bromeliaceae family (Ae. bromeliifolia (Rudge) Baker and Ananas comosus (L.) Merr.) where offered in combination with Ae. bracteata in two-choice bioassays, as in the previous experiment. Thirty replicates were performed for each comparison with workers originating from colonies living in bromeliads, and 21 to 26 replicates with ants from colonies collected in tree cavities. For each replicate, the ants had the choice between two shelters, i.e., one shelter made from the leaves of Ae. bracteata (control), and another made from the leaves of one of the two other bromeliad species.

Experiment Three
To evaluate the influence of other bromeliad-related traits (structure of the plant), ants were offered the choice between whole plants of similar size of the two Aechmea species. Thirty replicates were performed with both types of workers.

Experiment Four
As the results of the previous experiments showed that both shelters made of Ae. bracteata leaves and whole plants of this species were preferred (see Results), we evaluated whether the choice by N. villosa ants could be influenced by the size of the available Ae. bracteata. Twenty-nine replicates with workers from colonies nesting in Ae. bracteata were set up. In each replicate, Ae. bracteata bromeliads of two different sizes were offered, i.e., small (25 cm) vs. large (80 cm).

Statistical Analyses
Not all ants were found inside refuges; some workers were foraging, and others were inactive. Inactivity is very common in social insects and is an intrinsic feature of the ants' behavior [68], making up to 40% of the members in a colony (e.g., [69]); furthermore, specific workers are consistently inactive [70]. Inactive workers are quantitatively important in N. villosa colonies [19]. To avoid inconsistencies due to a number of ants not choosing or performing other activities, we calculated the total number of ants found inside the proposed refuges or plants and used proportions of ants as the variable response.
Data (successes and failures) were analyzed fitting a generalized linear mixed model (GLMM) with a binomial error distribution and a logit link (maximum likelihood). To control for any bias due to colony-level effects, colony was included as a random effect, and the treatments (with or without experience with Ae. bracteata) as a fixed effect. To explore the magnitude of the fixed effect, we performed a likelihood ratio test (Wald Chi-square test); 95% CI allowed to infer differences within treatments. Analyses were performed in package LME4 in R version 3.6.2 [71,72].

Ethics Statement
Sampling complied with the current laws of Mexico and was carried out under permit number FAUT-0277 from Secretaría de Medio Ambiente y Recursos Naturales, Dirección General de Vida Silvestre, granted to G.P.-L. Only the biological material required for this study was collected.

Species Identification
DNA sequences generated in the present work (Genbank Accession Numbers MK779595 to MK779604) confirmed that both populations (ants nesting in Ae. bracteata and ants nesting in live tree cavities) did not diverge genetically. All ants used in this work belonged to N. villosa. DNA sequences of both populations represent a single molecular operational taxonomic unit, and cluster with all other N. villosa molecular public data ( Figure S1).

Tandem Running Behavior
In the bioassays, N. villosa workers began exploring both the new area (election chamber) and the proposed refuges, and then selected one shelter. Afterwards, workers returned to the nest box and recruited nestmates through tandem running behavior. These recruited workers moved to the selected refuge, inspected it, and returned to the "old nest" to recruit new nestmates. Qualitatively, we found that several tandem running ants followed the same path, suggesting trail laying behavior, although marking of the trail was not observed (Video S1). Similar trail laying through hind gut fluids or pygidial gland secretions has been reported in other ponerine species using tandem running [73][74][75][76].
The recruitment process was initiated through a "jerking" movement of a recruiting ant stimulating a nestmate to follow her to the new nest site, whereby the ant performs a rapid and vertical shaking of the body. Such rapid, vertical shaking of the body displayed by the recruiting ant to enhance the chemical signal has been reported in other ant species from various subfamilies [77,78]. Nestmates reacted by replicating the jerking movements, and then initiated tandem running along the trail of the recruiting ant toward the new nest site. A single worker, or occasionally a maximum of two, were recruited and travelled in a single column. In some cases, when contact between the scout and the follower was broken, the recruiting ant pulled the legs of the nestmate with their jaws to reinforce the recruitment signal.

Experiment One
Ants from both origins (whether originally nesting in bromeliads or in tree cavities) significantly preferred refuges made up of the leaves of Ae. bracteata over cardboard shelters: 0.978 ± 0.0002 (mean proportion ± SEM, 95% CI [0.9774-0.9781]) for ants originally nesting in Ae. bracteata and 0.989 ± 0.0001 (95% CI [0.9892-0.9897]) for those nesting in tree cavities ( Figure 1). However, the origin of ants had a significant influence on their choices (GLMM, Wald χ 2 = 8338.5, df = 1, p < 0.0001) as ants originally nesting in bromeliads had a significantly lower probability of choosing the refuge made up of leaves of the bromeliad than ants originally nesting in tree cavities. In general, the brood was transported to the shelters with the higher proportion of workers.
Insects 2020, 11, x 6 of 16 stimulating a nestmate to follow her to the new nest site, whereby the ant performs a rapid and vertical shaking of the body. Such rapid, vertical shaking of the body displayed by the recruiting ant to enhance the chemical signal has been reported in other ant species from various subfamilies [77,78]. Nestmates reacted by replicating the jerking movements, and then initiated tandem running along the trail of the recruiting ant toward the new nest site. A single worker, or occasionally a maximum of two, were recruited and travelled in a single column. In some cases, when contact between the scout and the follower was broken, the recruiting ant pulled the legs of the nestmate with their jaws to reinforce the recruitment signal.

Experiment One
Ants from both origins (whether originally nesting in bromeliads or in tree cavities) significantly preferred refuges made up of the leaves of Ae. bracteata over cardboard shelters: 0.978 ± 0.0002 (mean proportion ± SEM, 95% CI [0.9774-0.9781]) for ants originally nesting in Ae. bracteata and 0.989 ± 0.0001 (95% CI [0.9892-0.9897]) for those nesting in tree cavities ( Figure 1). However, the origin of ants had a significant influence on their choices (GLMM, Wald χ 2 = 8338.5, df =1, p ˂ 0.0001) as ants originally nesting in bromeliads had a significantly lower probability of choosing the refuge made up of leaves of the bromeliad than ants originally nesting in tree cavities. In general, the brood was transported to the shelters with the higher proportion of workers. . Workers from colonies of two distinct origins were tested, i.e., ants originally nesting in Ae. bracteata (n = 30 trials, grey bars) and ants from colonies previously nesting in tree cavities (n = 30 trials, white bars). * p ˂ 0.05, *** p ˂ 0.0001, GLMM Wald χ 2 .

Experiment Two
When ants had to select between refuges made up of leaves of two Aechmea species, their choice was significantly influenced by their previous nesting site (GLMM, Wald χ 2 = 13.067, p = 0.00035). Ants from tree cavities that did not have any previous contact with Ae. bracteata have a higher probability of choosing Ae. bracteata over Ae. bromeliifolia (0.960 ± 0.027, 95% CI [0.8592-0.9895]) than ants from colonies originally nesting in Ae. bracteata (0.581 ± 0.089, 95% CI [0.4043-0.7391]). Ants from colonies nesting in Ae. bracteata did not show any marked preference ( Figure 2). However, when the choice concerned refuges made up of leaves of species from two different Bromeliaceae genera, ants from both origins preferred Ae. bracteata over A. comosus (ants nesting in bromeliads: 0.952 ± 0.018, 95% CI [0.9015-0.9776]; ants in tree cavities: 0.929 ± 0.033, 95% CI [0.8296-0.9720]) ( Figure 2); and although this pattern was stronger in ants originally nesting in Ae. bracteata, the probability of choosing between Ae. bracteata and A. comosus was not influenced by the origin of ants (GLMM, Wald χ 2 = 0.486, p = 0.48).

Experiment Three
The origin of the ants had no effect on the probability of choosing between whole plants of Ae. bracteata  . Workers from colonies of two distinct origins were tested, i.e., workers originally nesting in Ae. bracteata (n = 30 trials, grey bars) and workers from colonies previously nesting in tree cavities (n = 21 or n = 26 trials, white bars). * p < 0.05, *** p < 0.0001, GLMM Wald χ 2 .

Experiment Three
The origin of the ants had no effect on the probability of choosing between whole plants of Ae. bracteata and Ae. bromeliifolia (GLMM, Wald χ 2 = 0.011, p = 0.916). Ants of both origins preferred Ae. bracteata over Ae. bromeliifolia plants (ants originally nesting in Ae. bracteata: 0.661 ± 0.034, 95% CI

Discussion
Most animals, if not all, exhibit innate behaviors in response to specific sensory stimuli [79,80]. Bumble bees and honeybees, for example, exhibit innate color preferences, notably yellow and purple, which reflect the peak sensitivity of their color receptors [81][82][83][84], while the pseudomyrmecine ants Tetraponera aethiops Smith show an innate attraction to their natural host plant Barteria fistulosa Mast. (Passifloraceae), even when young callows were reared in laboratory conditions without any further contact with this plant post emergence [37]. Our experiments provide a similar example of innate attraction of N. villosa workers towards the bromeliad Ae. bracteata. Whatever the origin of the ants (with or without previous experience with Ae. bracteata), our data show that N. villosa workers exhibit a significant preference for refuges made up of leaves of Ae. bracteata over other available potential refuges consisting of leaves of another bromeliad species, A. comosus, or of cardboard; however, ants with a previous experience with Ae. bracteata did not differentiate between refuges made up of leaves of this bromeliad and those of the close Ae. bromeliifolia. Contrary to the results obtained by [25], which did not demonstrate any spontaneous preference of N. villosa for Ae. bracteata (workers reared in the laboratory without any contact with plants were attracted indifferently towards Ae. bracteata or towards the orchid Myrmecophila tibicinis (Batem.) Rolfe), our results point to the existence of an innate preference for Ae. bracteata. In all of the bioassays, workers without previous experience with this bromeliad significantly preferred Ae. bracteata, although a proportion of ants did not engage in emigrations (inactive ants).
The influence of the environment, through preimaginal and neonatal learning (early experience), can interfere and replace any innate attraction or repulsion [85,86]. In ants, environmental induction of adult choices by passive familiarization during early adulthood has been demonstrated for various species. For example, although under natural conditions thyme (Thymus vulgaris L.) repels adult workers of the formicines Formica polyctena Foerster and Camponotus vagus Scopoli, workers of these species chose to settle in tubes that contained this plant if they have been reared in its presence when they were callow neonates [35,87]. Similarly, in the African arboreal ants T. aculeatum and O. longinoda, early learning during the first part of the life of adult workers and gynes can supersede an innate attraction to guava (Psidium guajava L.) and cocoa (Theobroma cacao L.) or to mango leaves (Mangifera indica L.), respectively [36]. Early learning, during the larval life and/or just after adult emergence, appears to occur to some extent in N. villosa, as our data show that the preference for Ae. bracteata was modulated by the original nesting substrate; previous experience with Ae. bracteata enhancing the preference towards this plant over cardboard or A. comosus. These results confirm previous studies on foundresses of N. villosa which have shown that the influence of the original nest site environment on subsequent individual choice during nest site selection for colony foundation is due to imprinting during larval life, strengthened at the beginning of the imago life through early learning [25]. When given the choice between a refuge containing Ae. bracteata or nothing, gynes of N. villosa that have previously experienced contact with the bromeliad during their larval life and the first part of their pupal stage significantly preferred Ae. bracteata, whereas gynes which completed the pupal stage on another epiphyte (M. tibicinis) did not discriminate between both options. However, gynes which completed their pupal life on M. tibicinis displayed a significant preference to this plant, when presented with a choice between Ae. bracteata and M. tibicinis [25]. The local fidelity towards Ae. bracteata over other available nest sites observed in our study area and the marked preference for this epiphyte during our experiments could be explained through early learning by winged queens, as suggested by [25], combined with an innate attraction to Ae. bracteata in workers, influencing their choice during nest relocation events. The evolution of such an innate attraction, leading to local specialization in Ae. bracteata as a nesting site, could be due to the predominance of this particular bromeliad throughout the biogeographic area of N. villosa (Ae. bracteata occurs from E Mexico to N Colombia and NW Venezuela [88]), and also because this bromeliad as a microhabitat provides specific benefits to ants, both as a shelter and as a foraging site, and further constitutes a refuge during extreme stochastic climatic events [33,89]. Preference of N. villosa workers for Ae. bracteata could be an adaptive response driven by climatic events (droughts, floods) in our study area, promoting the selection of such a stable microhabitat.
Various other bromeliad associated traits (size, chemical profile) could also be involved in nest site selection as demonstrated by the preference of N. villosa workers for large Ae. bracteata when presented with small and large plants, or by the differences in preference demonstrated when choosing between closely related bromeliad species (both when offered as refuges and complete plants), most likely related to the chemical substances they produce. The bromeliads used in this study share similar traits (e.g., long and narrow leaves), and the texture of their leaves and general architecture are similar, but not identical; furthermore, the nature and composition of their chemical signals are different, particularly between species from different genera [90], and it is possible that ants use any small difference in structure or composition of the leaves during nest site selection. In addition, the architectural form and complexity of the plant could facilitate the emergence of different temperature and humidity microhabitats suitable for ants to settle, and it is known that other specialist arthropods (e.g., bromeliad-dwelling salticids) choose bromeliads based on rosette and leaf architectures [91]; furthermore, larger individual plants with a complex structure facilitate the development of large ant populations and promote the maintenance of a diverse community of potential prey [92]. The use of chemical cues for host localization has been reported in a wide range of insects, including both herbivores [93][94][95][96] and predators or parasites [97][98][99]. In ants, chemical volatiles are also used to identify potential host plants. For example, Crematogaster spp. foundress queens can recognize their Macaranga host plant species, identifying chemical compounds of the stem surfaces of seedlings [46]; and queens of Azteca spp. and Allomerus octoarticulatus Mayr use chemical cues to select their myrmecophyte Cordia nodosa Lam. [47]. In our experiments, it was evident that N. villosa workers, originally nesting in Ae. bracteata, had difficulty in arriving at a consensus when in the presence of shelters of two Aechmea species, without further information on the suitability of the potential nest sites (plant structure or size of the cavity, for example), and therefore the probabilities of choosing either bromeliad were similar ( Figure 2). Contrastingly, choosing between complete plants of these two bromeliads was straight forward and significantly in favor of Ae. bracteata over the other species, although some ants also settled inside Ae. bromeliifolia plants. Noteworthy, a proportion of ants did not engage in emigrations (inactive ants). Our failure to show any statistical difference when ants chose between refuges of these two bromeliads could have arisen, in part, due to the presence of inactive workers and the possible exclusion of specialized workers in our reduced experimental groups. This could have influenced the number of ants in shelters after 24 h, as a significant higher latency in performing emigration tasks has been reported for workers not specialized on emigration tasks in Temnothorax ants [65]. Considered together, these results suggest that workers of N. villosa select Ae. bracteata through some plant stimuli, probably of chemical nature, supplemented with information provided by the whole plant. The ability to discriminate between distinct plant species and genera is an obvious advantage, as the time and energy to find a suitable nest site is minimized.
Evidently, nearly all ant species have the capability of shifting their nests if they become unsuitable [1] and selecting the best nest site among numerous alternatives can be critical to the success and survival of the colony. As our results show, in most instances, N. villosa workers preferred to settle in Ae. bracteata shelters and plants over other possibilities, and preferred large Aechmea plants over small ones. Behavioral flexibility constitutes an essential component of the adaptive repertoires of animals. In this context, it is not impossible that modulation of the innate attraction of N. villosa workers towards Ae. bracteata through early experience could facilitate the replacement of this plant as the most suitable nest site in habitats where Ae. bracteata is rare, as occurs in the northern part of the Yucatan Peninsula where this species nests in cavities of several live trees.

Conclusions
In our experiments, nest site choice by N. villosa workers was influenced by an innate attraction to the bromeliad Ae. bracteata. The local fidelity towards Ae. bracteata over other available nest sites in the southern region of the Yucatan Peninsula and the marked preference for this epiphyte in our experiments could be explained through this innate attraction to Ae. bracteata influencing the choice of workers during nest relocation events. Reinforcement of this preference by preimaginal and early learning during adulthood is not excluded, but more experiments are needed specifically targeting early learning in this species. Preference of N. villosa workers for Ae. bracteata could be an adaptive response driven by extreme climatic events in our study area, promoting the selection of such a stable microhabitat as a nesting site.
Supplementary Materials: The following are available online at http://www.mdpi.com/2075-4450/11/3/200/s1, Table S1: Original composition of the Neoponera villosa colonies used in the two-choice bioassays, Figure S1: Taxon ID tree of N. villosa molecular public data, including the 10 sequences from this study (highlighted in yellow), Video S1: Characteristic recruitment behavior in Neoponera villosa. Note that two tandem pairs are following the same path, suggesting the existence of chemical trail laying.